From: Thomas Gleixner <tglx@linutronix.de>
Date: Sat, 1 Apr 2017 12:51:01 +0200
Subject: rtmutex: Provide locked slowpath
Git-repo: git://git.kernel.org/pub/scm/linux/kernel/git/rt/linux-rt-devel.git
Git-commit: 9b483495aea4175912581bd24f664a86aa1c81eb
Patch-mainline: Queued in subsystem maintainer repository
References: SLE Realtime Extension

The new rt rwsem implementation needs rtmutex::wait_lock to protect struct
rw_semaphore. Dropping the lock and reaquiring it for locking the rtmutex
would open a race window.

Split out the inner workings of the locked slowpath so it can be called with
wait_lock held.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Mike Galbraith <mgalbraith@suse.de>
---
 kernel/locking/rtmutex.c        |   72 +++++++++++++++++++++++-----------------
 kernel/locking/rtmutex_common.h |    8 ++++
 2 files changed, 50 insertions(+), 30 deletions(-)

--- a/kernel/locking/rtmutex.c
+++ b/kernel/locking/rtmutex.c
@@ -1739,30 +1739,13 @@ static void ww_mutex_account_lock(struct
 }
 #endif
 
-/*
- * Slow path lock function:
- */
-static int __sched
-rt_mutex_slowlock(struct rt_mutex *lock, int state,
-		  struct hrtimer_sleeper *timeout,
-		  enum rtmutex_chainwalk chwalk,
-		  struct ww_acquire_ctx *ww_ctx)
+int __sched rt_mutex_slowlock_locked(struct rt_mutex *lock, int state,
+				     struct hrtimer_sleeper *timeout,
+				     enum rtmutex_chainwalk chwalk,
+				     struct ww_acquire_ctx *ww_ctx,
+				     struct rt_mutex_waiter *waiter)
 {
-	struct rt_mutex_waiter waiter;
-	unsigned long flags;
-	int ret = 0;
-
-	rt_mutex_init_waiter(&waiter, false);
-
-	/*
-	 * Technically we could use raw_spin_[un]lock_irq() here, but this can
-	 * be called in early boot if the cmpxchg() fast path is disabled
-	 * (debug, no architecture support). In this case we will acquire the
-	 * rtmutex with lock->wait_lock held. But we cannot unconditionally
-	 * enable interrupts in that early boot case. So we need to use the
-	 * irqsave/restore variants.
-	 */
-	raw_spin_lock_irqsave(&lock->wait_lock, flags);
+	int ret;
 
 #ifdef CONFIG_PREEMPT_RT_FULL
 	if (ww_ctx) {
@@ -1778,7 +1761,6 @@ rt_mutex_slowlock(struct rt_mutex *lock,
 	if (try_to_take_rt_mutex(lock, current, NULL)) {
 		if (ww_ctx)
 			ww_mutex_account_lock(lock, ww_ctx);
-		raw_spin_unlock_irqrestore(&lock->wait_lock, flags);
 		return 0;
 	}
 
@@ -1788,13 +1770,13 @@ rt_mutex_slowlock(struct rt_mutex *lock,
 	if (unlikely(timeout))
 		hrtimer_start_expires(&timeout->timer, HRTIMER_MODE_ABS);
 
-	ret = task_blocks_on_rt_mutex(lock, &waiter, current, chwalk);
+	ret = task_blocks_on_rt_mutex(lock, waiter, current, chwalk);
 
-	if (likely(!ret))
+	if (likely(!ret)) {
 		/* sleep on the mutex */
-		ret = __rt_mutex_slowlock(lock, state, timeout, &waiter,
+		ret = __rt_mutex_slowlock(lock, state, timeout, waiter,
 					  ww_ctx);
-	else if (ww_ctx) {
+	} else if (ww_ctx) {
 		/* ww_mutex received EDEADLK, let it become EALREADY */
 		ret = __mutex_lock_check_stamp(lock, ww_ctx);
 		BUG_ON(!ret);
@@ -1803,10 +1785,10 @@ rt_mutex_slowlock(struct rt_mutex *lock,
 	if (unlikely(ret)) {
 		__set_current_state(TASK_RUNNING);
 		if (rt_mutex_has_waiters(lock))
-			remove_waiter(lock, &waiter);
+			remove_waiter(lock, waiter);
 		/* ww_mutex want to report EDEADLK/EALREADY, let them */
 		if (!ww_ctx)
-			rt_mutex_handle_deadlock(ret, chwalk, &waiter);
+			rt_mutex_handle_deadlock(ret, chwalk, waiter);
 	} else if (ww_ctx) {
 		ww_mutex_account_lock(lock, ww_ctx);
 	}
@@ -1816,6 +1798,36 @@ rt_mutex_slowlock(struct rt_mutex *lock,
 	 * unconditionally. We might have to fix that up.
 	 */
 	fixup_rt_mutex_waiters(lock);
+	return ret;
+}
+
+/*
+ * Slow path lock function:
+ */
+static int __sched
+rt_mutex_slowlock(struct rt_mutex *lock, int state,
+		  struct hrtimer_sleeper *timeout,
+		  enum rtmutex_chainwalk chwalk,
+		  struct ww_acquire_ctx *ww_ctx)
+{
+	struct rt_mutex_waiter waiter;
+	unsigned long flags;
+	int ret = 0;
+
+	rt_mutex_init_waiter(&waiter, false);
+
+	/*
+	 * Technically we could use raw_spin_[un]lock_irq() here, but this can
+	 * be called in early boot if the cmpxchg() fast path is disabled
+	 * (debug, no architecture support). In this case we will acquire the
+	 * rtmutex with lock->wait_lock held. But we cannot unconditionally
+	 * enable interrupts in that early boot case. So we need to use the
+	 * irqsave/restore variants.
+	 */
+	raw_spin_lock_irqsave(&lock->wait_lock, flags);
+
+	ret = rt_mutex_slowlock_locked(lock, state, timeout, chwalk, ww_ctx,
+				       &waiter);
 
 	raw_spin_unlock_irqrestore(&lock->wait_lock, flags);
 
--- a/kernel/locking/rtmutex_common.h
+++ b/kernel/locking/rtmutex_common.h
@@ -131,6 +131,14 @@ extern bool __rt_mutex_futex_unlock(stru
 
 extern void rt_mutex_postunlock(struct wake_q_head *wake_q,
 				struct wake_q_head *wake_sleeper_q);
+/* RW semaphore special interface */
+struct ww_acquire_ctx;
+
+int __sched rt_mutex_slowlock_locked(struct rt_mutex *lock, int state,
+				     struct hrtimer_sleeper *timeout,
+				     enum rtmutex_chainwalk chwalk,
+				     struct ww_acquire_ctx *ww_ctx,
+				     struct rt_mutex_waiter *waiter);
 
 #ifdef CONFIG_DEBUG_RT_MUTEXES
 # include "rtmutex-debug.h"